Dynamic Load Balancing of Black-Box Applications with a Resource Selection Mechanism on Heterogeneous Resources of the Grid

Job and Application-Level Scheduling: An Integrated Approach for Achieving Quality of Service in Distributed Computing

“…The dynamic configuration of the environment, large number of resource reallocation events, user's and resource owner's needs as well as virtual organization policy of resource assignment should be taken into account. The scheduling strategy is formed on a basis of formalized efficiency criteria, which sufficiently allow reflecting economical principles [14] of resource allocation by using relevant cost functions and solving the load balance problem for heterogeneous processor nodes [2,13]. The strategy is built by using methods of dynamic programming [22] in a way that allows optimizing scheduling and resource allocation for a set of tasks, comprising the compound job.…”

“…First one is based on the usage of the available and nondedicated resources, where resource brokers are acting as agents between users and processor nodes [1,2]. Several projects such as AppLeS [3], APST [4], Legion [5], DRM [6], Condor-G [7], Nimrod/G [8] and others, which follow this idea, are often associated with application-level scheduling.…”

“…Distinct from existing scheduling solutions in distributed computing [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15]; our approach supposes techniques of dynamic redistribution of job-flows between processor nodes in conjunction with application-level scheduling. It is considered, that the job can be compound (multiprocessor) and the tasks, included in the job, are heterogeneous in terms of computation volume and resource need.…”

“…What sets our work apart from other scheduling research is that we consider integrated application-level and job-flow management as a fundamental part of the effective scheduling strategy within the virtual organization. The choice of the strategy depends on the utilization state of processor nodes [2,13], data storage and replication policies [10,17,18], the job structure (computational granularity and data dependencies), user estimations of completion time, resource requirements, and advance reservations [15,16,19]. Diverse application-level strategies are described in [20][21][22].…”

“…Another trend is based on the concept of virtual organizations and is mainly aimed at Grid systems [9][10][11]. Resource brokers, that are used in the first trend [1][2][3][4][5][6][7][8] are scalable and flexible and can be adapted to a specific application. Non-clustered Grid resource computing environments are using similar approach.…”